From access to parity — Can the DMA secure effective interoperability for on-device AI agents? Theory and evidence from the mobile AI stack

(2026)

Files

Fonck_52232100_2026.pdf
  • Open access
  • Adobe PDF
  • 972.81 KB

Details

Supervisors
Faculty
Degree label
Abstract
The Digital Markets Act (DMA) was conceived before the rise of large language models (LLMS). It is therefore worth asking whether its obligations reach the new competitive bottleneck that AI agents have created on the device itself. Taking the mobile market as its focal case, this thesis asks whether Article 6(7) DMA on interoperability, with Article 6(5) DMA on self-preferencing, effectively reaches a gatekeeper’s privileged access to the on-device AI accelerator (the neural processing unit) for its own agent, or whether the obligation’s access-not-parity scope, its integrity exception, and its enforcement mode leave a hardware-layer foreclosure gap. It combines a doctrinal reading of Article 6(7) and its enforcement, a reduced-form model of foreclosure by access-quality degradation, and a case study of the Commission’s Google proceeding (Case DMA.100220) and of Apple’s June 2026 decision to withhold its rebuilt Siri from iPhones and iPads in the European Union.